Low-illumination unmanned aerial vehicle image enhancement method and device based on structure and illumination dual guidance

By employing a dual-guided approach of illumination and structure using the ISDG-Net network, the problems of inaccurate illumination distribution estimation and loss of details of small targets in low-light UAV images are solved, achieving efficient illumination enhancement and detail restoration, thereby improving image quality and target detection accuracy.

CN122391041APending Publication Date: 2026-07-1410TH RES INST OF CETC
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
10TH RES INST OF CETC
Filing Date
2026-03-11
Publication Date
2026-07-14

AI Technical Summary

Technical Problem

Existing low-light drone image enhancement technologies suffer from inaccurate illumination distribution estimation and loss of detail in small target structures, leading to decreased image quality and reduced accuracy in subsequent computer vision tasks.

Method used

An ISDG-Net network based on dual structure and illumination guidance is adopted. Illumination priors and structure priors are generated through illumination estimation branch and structure preservation branch, and feature fusion is performed in combination with the backbone network to achieve illumination enhancement.

Benefits of technology

It improves the efficiency of illumination calculation, maintains the natural transition of light and shadow in the image, restores the details of small targets, and enhances the image quality and the recognition accuracy of subsequent target detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122391041A_ABST
    Figure CN122391041A_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of computer vision and image processing, and provides a low-illumination unmanned aerial vehicle image enhancement method and device based on structure and illumination double guidance, which comprises the following steps: acquiring a low-illumination image; inputting the low-illumination image into an ISDG-Net network for illumination enhancement; the ISDG-Net network comprises an illumination estimation branch, a structure reservation branch and a backbone network; wherein the low-illumination image is input into the illumination estimation branch, the structure reservation branch and the backbone network respectively; the illumination estimation branch generates illumination priori; the structure reservation branch generates structure priori; the backbone network performs feature fusion double guidance enhancement based on the illumination priori and the structure priori to generate an illumination enhanced image. The present application can clearly restore dark ground object details and solve the problems of inaccurate illumination distribution estimation and loss of micro-target structure details in the existing low-illumination unmanned aerial vehicle image enhancement technology.
Need to check novelty before this filing date? Find Prior Art